Geographic Adjustment of Medicare Payments to Physicians ... Adjustment of Medicare Payments to...

Geographic Adjustment of Medicare Payments to Physicians: Evaluation of IOM Recommendations

July 2012

Thomas MaCurdy Jason Shafrin Thomas DeLeire Jed DeVaro Mallory Bounds David Pham Arthur Chia

Acumen, LLC

500 Airport Blvd., Suite 365

Burlingame, CA 94010

[This page is intentionally left blank.]

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians

EXECUTIVE SUMMARY

Medicare pays physicians for their services according to the Physician Fee Schedule

(PFS), which specifies a set of allowable procedures and payments for each service. Each

procedure is interpreted as being produced by a combination of three categories of inputs:

physician work (PW), practice expense (PE), and malpractice insurance (MP). The particular

blend of PW, PE, and MP inputs assessed to produce a service specifies its composition of

relative value units (RVUs). A payment for a procedure depends on its assigned RVUs and the

input prices assessed for each RVU component. Under mandates in Section 1848(e) of the

Social Security Act, the Centers for Medicare and Medicaid Services (CMS) must apply

geographic cost indices in the calculation of component RVU input prices. In 1992, CMS

introduced Geographic Practice Cost Indices (GPCIs) to comply with this mandate; CMS

updates GPCIs at least every three years.

In its latest efforts to improve the methodology and data sources used to compute GPCIs

and other geographic input cost adjustments, CMS funded an Institute of Medicine (IOM) study

to identify areas where the GPCI methodology could be improved. In its 2011 Phase I report,

IOM evaluates the methodology CMS uses to make adjustments to the PFS and the extent to

which alternative sources of data are representative of the economic circumstances healthcare

providers face. The IOM study also offers a number of proposed modifications to the

methodology CMS uses to compute GPCI values. This report evaluates IOM’s recommended

changes to the GPCI methodology.

How GPCIs Affect Physician Payments

GPCIs measure geographic differences in input prices. Paralleling the RVU structure,

GPCIs are split into three parts: PW, PE, and MP. Each of these three GPCIs adjusts its

corresponding RVU component. GPCIs do not affect aggregate payment levels; instead, they

reallocate payment rates to reflect regional variation in relative input prices. For example, a PE

GPCI of 1.2 indicates that practice expenses in that area are 20 percent above the national

average, whereas a PE GPCI of 0.8 indicates that practice expenses in that area are 20 percent

below the national average. CMS calculates the three GPCIs for payment areas known as

Medicare localities. Each physician payment locality is assigned an index value, which equals

the area’s estimated input cost divided by the average input cost nationally. Localities are

defined alternatively by state boundaries (e.g., Wisconsin), metropolitan statistical areas (MSAs)

(e.g., Metropolitan St. Louis, MO), portions of an MSA (e.g., Manhattan), or rest-of-state areas

that exclude metropolitan areas (e.g., Rest of Missouri). As a result, some localities are large

metropolitan areas, such as San Francisco and Boston, whereas many localities are statewide

payment areas that include both metropolitan and nonmetropolitan areas, such as Minnesota,

Ohio, and Virginia.

i

Across these localities, CMS uses the conversion factor (CF), to calculate the payment

for each service in dollars. The conversion factor, which is updated annually, indicates the dollar

value CMS assigns to an RVU. The equation below demonstrates how CMS combines the CF

with the PW, PE, and MP GPCIs and the corresponding RVUs to establish a Medicare physician

payment for any service H in locality L:

CMS calculates GPCIs using six component indices. Whereas the PW and MP GPCIs

are based on a single component index, the PE GPCI is comprised of four component indices

(i.e., the employee wage; purchased services; office rent; and equipment, supplies and other

indices). The PE GPCI is calculated as a weighted average of the four PE GPCI component

indices, where the weight assigned to each PE GPCI component index equals each input’s

average share of physician practice expenses nationally. Table 1 below provides additional

information on each component index.

Table 1: Breakdown of GPCIs into Six Component Indices

GPCI Component Index Measures Geographic Differences in:

Physician

Work Single Component Physician wages

Practice

Expense

Employee Wage Wages of clinical and administrative office staff

Purchased Services Cost of contracted services (e.g., accounting, legal,

advertising, consulting, landscaping)

Office Rent Physician cost to rent office space

Equipment, Supplies, and Other Practice expenses for inputs such as chemicals and

rubber, telephone use and postage

Malpractice Single Component Cost of professional liability insurance

Although GPCIs affect payments for each procedure depending on the relative amounts

of PW, PE, and MP RVUs, one can summarize the overall impact of the GPCI components on a

locality’s physician reimbursement levels, using the Geographic Adjustment Factor (GAF). The

GAF is calculated as the weighted average of the three GPCIs, where the weights are the

percentage of RVUs nationally made up by the PW, PE, and MP RVUs. For calendar year (CY)

2012, one can calculate the GAF as follows:

Overview of IOM’s GPCI Recommendations

IOM recommended alterations of GPCIs fall into five broad categories shown in Table 2.

The first column lists the recommendation category, the second column identifies the

ii Executive Summary Acumen, LLC

recommendation numbering system from IOM’s report, and the third presents a brief description

of these recommendations. Whereas the first three recommendation categories propose changes

to the current GCPI methodology, the latter two endorse aspects of the current CMS approach.

This report focusses on evaluating the potential impacts of the first three categories of IOM

recommendations that propose revisions to the current methods for calculating GPCIs.

Table 2: IOM Geographic Practice Cost Index (GPCI) Recommendations

Category Number Description

Employee

Wages

2-1

The same labor market definition should be used for both the hospital wage index

and the physician geographic adjustment factor. Metropolitan statistical areas

and statewide non-metropolitan statistical areas should serve as the basis for

defining these labor markets.

2-2 The data used to construct the hospital wage index and the physician geographic

adjustment factor should come from all healthcare employers.

4-1

Wage indexes should be adjusted using formulas based on commuting patterns

for healthcare workers who reside in a county located in one labor market but

commute to work in a county located in another labor market.

5-4 The practice expense GPCI should be constructed with the range of occupations

employed in physicians’ offices, each with a fixed national weight based on the

hours of each occupation employed in physicians’ offices nationwide.

5-5 CMS and BLS should develop a data use agreement allowing the Bureau of

Labor Statistics to analyze confidential BLS data for CMS.

Physician

Wages

5-2 Proxies should continue to be used to measure geographic variation in the

physician work adjustment, but CMS should determine whether the seven proxies

currently in use should be modified.

5-3 CMS should consider an alternative method for setting the percentage of the work

adjustment based on a systematic empirical process.

Office Rent 5-6 A new source of data should be developed to determine the variation in the price

of commercial office rent per square foot.

Purchased

Services 5-7 Nonclinical labor-related expenses currently included under PE office expenses

should be geographically adjusted as part of the wage component of the PE.

Cost Share

Weights 5-1

GPCI cost share weights for adjusting fee-for-service payments to practitioners

should continue to be national, including the three GPCIs (work, practice

expense, and liability insurance) and the categories within the practice expense

(office rent and personnel).

Although not to become a part of IOM’s formal recommendations until its Phase II

report, a theme guiding recommendations throughout IOM’s Phase I report is the development of

a three-tiered system for defining payment areas: the first tier consists of counties to be used as

the basis for calculating employee wage indices with adjustments incorporated to account for

workers’ commuting patterns across MSAs; the second tier comprises MSA-type areas to be

used for the geographic cost adjustments of PE GPCI components such as office rents, purchased

services, and malpractice insurance; and the third tier consists of a national payment area for PE

GPCI items as "Equipment, Supplies and Other." Table 3 presents an overview of IOM’s

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians iii

suggested replacements of current GPCI localities by payment areas tailored to capture the

market environments appropriate for determining payment of individual GPCI components. The

rows of this table list the six individual GPCI components incorporated in the PFS and the

columns list the regions entertained as candidates for calculating geographic adjustments of

payments to physicians. Readers may know the "statewide tier" payment area, which combines

counties into tiers within each state based on each county’s GAF value, as the "Option 3"

payment area definition presented in the July 2007 proposed rule. Returning to the table, an "X"

in a row indicates that the payment area suggested by IOM to compute the GPCI component.

One sees in this table that IOM favors MSAs as the principal choice for payment areas, with

counties playing a role for employee wage indices and a national market for equipment and

supplies. The empirical analyses in later sections assess the impacts of considering each of the

payment area candidates listed in Table 3, with the goal of placing the IOM recommendations in

useful context.

Table 3: IOM’s Suggested Three-Tiered System for Defining GPCI Payment Areas

GPCI Expense Category

Payment Area

County MSA

Statewide

Tier Locality National

Physician Work X

Practice Expense

Employee Wage X

Purchased Services X

Office Rent X

Equipment, Supplies, Other X

Malpractice Insurance X

Evaluation of IOM Recommendations for the Employee Wage Index

IOM proposes two notable changes to the current employee wage index (EWI)

methodology. First, IOM recommends redefining the payment areas CMS uses to calculate EWI

values. Second, IOM proposes that CMS measure worker wages within these payment areas

using data limited to workers employed in the healthcare industry (rather than across all

industries).

IOM Recommendations to Redefine Payment Areas for the Employee Wage Index

IOM’s proposal for revising payment areas would permit EWI values to vary across

counties, including for counties located in the same MSA. If implemented, the number of EWI

payment areas would increase from 89 to potentially over 3,000. There exists substantial

variation in employment costs within each of the current 89 locality-based payment areas. To

iv Executive Summary Acumen, LLC

adjust for this variability, IOM suggests calculating wage rates based on MSA data and inferring

wage rates for counties through smoothing algorithms that account for computing patterns from

counties to MSAs. This recommendation for GPCI wage calculations matches that proposed by

IOM for the hospital wage index (HWI).

Four steps characterize IOM’s proposals for calculating EWI values for each physician

practice:

(1) Compute the mean/median hourly wage (MHW) for each MSA;

(2) Calculate an area index wage for each county based on out-commuting patterns;

(3) Assign an index wage to each physician office based on its county location; and

(4) Normalize physician office wage measures to create the employee wage index.

To illustrate these steps, consider a simple example shown in Table 4. In this example

there are two physician practices; Physician Office 1 is located in County A in MSA a, and

Physician Office 2 is located in County B in MSA b. Step 1 estimates the median/mean wage for

each MSA. This step essentially replicates the current employee wage index methodology, but

calculates a wage index value at the MSA rather than the locality level. Since this example only

has one physician office in each MSA, each MSA’s median wage equals the physician office

wage. The sixth column of Table 4 displays the MHW as calculated under step 1 for each MSA.

Table 4: Example Application of the IOM Out-Commuting Adjustment

Physician

Office

Physician

Office

Wage

Worker

County of

Residence

MSA where

Worker is

Employed

County-to-

MSA Out-

Commuting

Shares

Current EWI

Median

Hourly Wage

(Step 1)

IOM EWI

Commuting-

Adjusted

Index Wage

(Steps 2, 3)

1 $30 A a 80% $30 $28

b 20% $30 $28

2 $20 B a 20% $20 $22

b 80% $20 $22

Step 2 applies a commuting-based smoothing adjustment to create area index wages for

each county. Specifically, the county wage indices equal a weighted average of the MHW values

calculated in Step 1, where the weights are county-to-MSA out-commuting patterns. IOM’s out

commuting-based weights are defined as the share of workers who live in a county where the

physician office is located who commute out to work in a physician office in another MSA. This

modification differs from an in-commuting adjustment, which is based on the share of workers

who are employed at physician offices (or areas where offices are located) who commute from

other areas. The fifth column of Table 4 displays the county-to-MSA out-commuting shares, and

the seventh column presents each county’s commuting-adjusted area index wage. One can

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians v

calculate IOM EWI values for County A, for instance, as: $30×80% + $20×20% = $28; for

County B, the calculation is $30×20% + $20×80% = $22.

Step 3 sets each physician office’s wage measure equal to the Step 2 area wage of the

county in which the office is located. Because the out-commuting adjustment envisioned by

IOM in Step 2 varies by county, employee wage index values—and thus the PE GPCI as a

whole—also potentially vary by county depending on the smoothing option chosen.

Paralleling the current EWI methodology, Step 4 normalizes out-commuting-adjusted

wage measures by dividing each physician’s wage measure by the PE RVU-weighted average

wage measures for all offices. Although not shown in this example, this step produces an index

whose PE RVU-weighted average value equals 1.

Through the use of out-commuting shares to weight the wages of physician office

employees across MSAs, IOM’s proposal redefines the EWI to measure the wage levels

associated with the workers who live in a county rather than the workers who are employed in

the county. The purpose of a wage index, however, is to measure the earnings of healthcare

workers employed in a county, for this represents the costs of labor faced by the providers who

hire in the county. The relevant input price physician practices must pay to compete in their

pertinent labor market depends not only on the wage levels of individuals living nearby but also

on the wage levels paid to attract individuals living outside the local area who work at the

practices. As shown in this report, the values of the wage indices associated with healthcare

workers living in a county verses the workers employed in a county can be quite different.

Moreover, the IOM smoothing adjustment can produce counterintuitive EWI values,

especially in cases where a large share of workers commute from one MSA to another. Even if

all practices in a county pay their workers an identical wage, the IOM method increases these

practices’ EWI values above that wage if workers living in that county commute to MSAs where

practices pay higher wages. The reverse is true if workers living in this county commute to

MSAs where practices pay lower wages. Further, in the extreme case where all workers in a

county out-commute to another MSA, the EWI for physician practices in that county depends

entirely on the wage levels paid by practices located in other MSAs.

When IOM’s approach is applied in practice, this report concludes that IOM’s out-

commuting adjustment does reduce the size of cliffs. For counties in different localities that are

located within 50 miles of one another, applying the smoothing algorithm to the employee wage

index reduces the differences in GAF values by 0.14 percentage points (i.e., 0.0014) relative to

the MSA payment area definition without smoothing. Although the magnitude of this change is

small, recall that the IOM recommendation only applies the smoothing algorithm to the

employee wage index, and the employee wage index constitutes only 19 percent of the total GAF

value. Applying the smoothing methodology marginally reduces the frequency with which

vi Executive Summary Acumen, LLC

nearby counties have GAF differentials exceeding 5 percentage point. Thus, not only does the

average difference in GAF values decrease for counties located close to one another, but the

share of counties with large cliffs also decreases.

IOM Recommendations for Measuring Employee Wages

IOM’s proposal to measure wages for workers using data from the healthcare industry

rather than from all industries offers a number of conceptual advantages and disadvantages, but it

would likely have little effect on GAF values. An obvious attractive feature of such a change in

data sources relates to capturing geographic variation in worker wages that is idiosyncratic to

employment in the healthcare industry. On the other hand, the IOM approach has two

drawbacks. First, limiting the wage estimates to workers in the healthcare industry reduces the

sample size and thus decreases the precision of the wage estimates. This issue is particularly

relevant when measuring wages in sparsely populated rural areas. Second, measuring healthcare

industry wages across different geographic areas using BLS OES data requires access to

confidential BLS OES data, which may be difficult to acquire and would reduce the transparency

of the GPCI methodology as providers would not have access to these data. Nevertheless,

IOM’s own calculations indicate that the correlation between all-industry and healthcare industry

wages is over 0.99. Thus, despite certain conceptual arguments that favor calculating the

employee wage index using healthcare worker wage data, the impact on GAF values is likely

small in practice.

Evaluation of IOM Recommendations for Physician Work GPCI

Current policy methodology calculates the PW GPCI index following four steps:

(1) Select proxy occupations to include in the PW GPCI index and calculate an

occupation-specific county-level index for each county;

(2) Assign weights to each proxy-occupation index based on the occupation’s national

share of wage bill;

(3) Apply 25 percent adjustment through the 'inclusion factor' to dampen responsiveness

of the PW GPCI to regional variation in the proxy-occupation index; and

(4) Adjust values to ensure budget neutrality.

Table 5 summarizes the key changes in the above steps recommended by IOM. IOM’s principal

proposal consists of computing PW GPCI based on a familiar regression framework. Regarding

Step 1, IOM endorses continued use of proxy occupations to measure regional variation in

physician wages, but suggests selecting them based on the goodness-of-fit and predictive

information conveyed by regression estimation statistics. With respect to Step 2, IOM

recommends weighting each occupation according to the value of its estimated regression

coefficient. For Step 3, IOM proposes an inclusion factor equal to the sum of the regression

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians vii

coefficients on the proxy occupation variables. IOM’s Step 4 is identical to the status quo

approach.

Table 5: Summary of Changes to PW GPCI Components

PW GPCI

Component Current PW GPCI IOM’s Recommendations

Proxy Occupations Seven occupational groups intended to

measure wages for professional workers

Can use current or an alternative set of

proxy occupations

Occupation Weights National wage shares Correlation with physician wages

Inclusion Factor 25% Sum of regression model’s coefficients

for the proxy occupations variables

Budget Neutrality

Adjustment

Normalize index so that PW RVU-

weighted average PW GPCI equals 1.0

Normalize index so that PW RVU-

weighted average PW GPCI equals 1.0

Whereas the current construction of PW GPCI essentially relies on price index theory

familiar throughout the policy community to measure price (and wage) differences across

regions and over time, the IOM suggested approach creates an index based on the predicted

values from a regression. The regression estimates implicitly produce shares for occupations in

the index that correspond to no interpretable market basket. Instead, the coefficient estimates

reflect the degree of correlations between the price of one labor commodity and the prices of

others across regions. The coefficients cannot be interpreted as shares; any individual share

(coefficient) can be negative or greater than one; the empirical findings presented in this report

reveal many instances of both these cases.

While difficult to interpret IOM’s PW GPCI as characterizing a classic form of a wage

index, the IOM approach nevertheless has a straightforward statistical interpretation as a

prediction of the relative regional wages of physicians forecasted using the relative regional

wages of comparable occupations. Of course, if the wages of the group of occupations deemed

to be related to physicians shift uniformly across regions, then all wage indices produce the same

findings, since the form of weighting does not matter. However, when non-uniform shifts occur,

then the form of weighting effects the values of indices and one must select which form best

capture the phenomena of interest. From an economics perspective, a regression model that

relates wages in regional markets mimics a reduced form specification with coefficients that

summarize the impacts of a wide range of market factors determining wages, including

differences the relative supplies and demands of occupations across regions, regional variation in

the number of hours various occupations work, and composition of specialists in each area.

Notwithstanding, if one interprets the goal of the PW GPCI as principally predicting regional

differences in physician wages regardless of the sources of variation, then the IOM candidate

offers a popular statistical candidate.

viii Executive Summary Acumen, LLC

An empirical application of a variant of the IOM regression specification using BLS OES

data reveals the following findings:

All regression specifications produce a wide range of coefficient values, including a

large number of negative values;

The regressions produce few coefficients that are statistically significantly different

from zero;

The R-squared measure of fit for the various models varies from 0.19 to 0.65,

depending on the diversity and number of MSAs included as observations in the

regression; and

The estimated IOM inclusion factor is near zero or negative.

The last finding in this list highlights problems with IOM’s suggestion that one can

interpret the sum of the regression coefficients on proxy occupation wages as a measure of the

inclusion factor used in current GPCI policy. This sum directly corresponds to a transformed

correlation coefficient physicians’ relative regional wages and IOM’s composite occupation

wage index. Consequently, the "IOM inclusion factor" need not fall between zero and one as is

the case with the inclusion factor under currently policy. The IOM inclusion factor can be

negative; it can exceed one; and it can even equal zero. Such instances occur in the empirical

findings reported here.

Evaluation of IOM Recommendations for the Office Rent Index

The PE GPCI office rent index currently relies on residential rental data to estimate

physicians’ costs for commercial office space. Using such rental data as a proxy for commercial

rents is valid as long as residential rents are proportional to commercial rents across payment

areas. While such circumstances can occur in flexible markets where people can use land for

both residential and commercial purposes, markets can readily produce differential demands for

residential and commercial properties due to such factors as zoning laws. Additionally, both

demand and supply factors could cause geographic variation in residential rents to not be

proportional to regional variation in commercial rents. Due to the limitations of using residential

rent data, IOM proposes that a new source of data be developed to determine the variation in the

price of commercial office rent per square foot.

IOM’s proposal for identifying a source of commercial rent data to compute the office

rent index offers a number of attractive features. Although collecting rent data from physicians

could improve the accuracy of the office rent index, such an effort would encounter several

challenges: (i) collecting a new source of office rent data would be administratively costly, (ii)

physician response rates are typically low, (iii) utilizing office rent data collected directly from

physicians would introduce a circularity problem, and (iv) developing and collecting a new

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians ix

source of commercial office rent might partially replicate existing data sources currently being

studied. Our report identifies commercial rent data from the CoStar Group as a potential

candidate to replace the residential rent data currently used by GPCI in its calculations. CoStar

offers a detailed database that contains national commercial office rent data for over 2.8 million

commercial properties covering over 10 billion square feet of space. The database also tracks a

wide variety of property types and contains a relatively large number of commercial property

listings for rural states. The disadvantages of using CoStar are that it is fairly expensive and—

since the data source is proprietary—providers would not be able to fully validate the office rent

index calculations. This report recommends that future research should examine the impact of

using CoStar commercial rent data on the office rent index. Until these data are studied,

however, in the short-term this report recommends the continued use of the large and nationally

representative residential rent data available in the ACS.

Summary of Empirical Impact Analysis

To determine whether the IOM recommendations cause a meaningful change in

physician GAF values in practice, this report conducts a series of impact analyses of the IOM

recommendations. Table 6 presents these summary statistics. The first column lists the impact

analyses carried out in this report. The second column specifies the number of counties or

localities used to calculate GAF values. The third and fourth columns describe the median

change and absolute mean change. The remaining four columns present the distribution of

absolute GAF changes.

Table 6: Distribution of Changes in GAF for Impact Analyses

Proposed IOM

Modification

Total

Obs.

Median

Change

Abs.

Change

Mean

Distribution of Absolute

GAF Changes

0.00 to

0.01

0.01 to

0.05

0.05 to

0.10 > 0.10

Three-Tiered

Payment Areas

3223

Counties -0.025 0.028 14.2% 77.8% 7.3% 0.8%

Regression-Based

PW GPCI

(FP Specification)

89

Localities 0.007 0.029 24.8% 58.4% 16.8% 0%

Alternative Proxy Occ.,

Current PW GPCI

Methodology

89

Localities 0.000 0.004 96.6% 3.3% 0% 0%

The two IOM policy recommendations that induce the largest changes in GAF values

consist of modifying the definitions of GPCI payment area and using a regression-based

approach to calculate the PW GPCI. In both cases, the average change in GAF values is around

3 percentage points. Since IOM’s proposal only applies the out-commuting adjustment to the

x Executive Summary Acumen, LLC

employee wage index, the changes in county GAF values under the three-tiered payment area are

similar in magnitude to what occurs when redefining all GPCI component payment areas to

MSAs. Using an alternative set of proxy occupations to calculate PW GPCI values under the

current methodology leads to less than a half of a percentage point change in GAF values.

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians xi

.......................................................................................................................

...........................................................................................................................

..................

...........................................................................

........................................................................................

..............................................................................

..........................................................................

........................................................................

...............................................................................

....................................................................

............................................................................................

...............................................................

..................................................

................................................................

...............................................

...................................................................

....................................................................

......................................

.....................

.................................

.......................................................

...............................................

.............

......................................

...................................

.....................................................................

................................................

.........................

.....................

..............................

..............................................

..................

...................

................................................

............................

.........................................

..................................................

.........................................

...................................................

.........................................................................................

................................................................................................

............................................................

.......................................................

......................................................

........................................................................

TABLE OF CONTENTS

Executive Summary i 1 Introduction 1 2 Geographic Adjustments of Physician Fee Schedule Under Current Policy 4 2.1 How GPCIs Affect Physician Payments 4 2.2 GPCIs’ Six Component Indices 5 2.3 Current Policy for Calculating GPCIs 6

2.3.1 Physician Work GPCI Methodology 7 2.3.2 Practice Expense GPCI Methodology 9 2.3.3 Malpractice GPCI Methodology 11 2.3.4 Data Sources Used to Calculate GPCIs 11 2.3.5 Legislative Adjustments 12

3 Description of IOM’s GPCI Recommendations 14 3.1 Recommended Changes to the Employee Wage Index 16

3.1.1 Redefining Labor Market Payment Areas 16 3.1.2 IOM Employee Wage Index: A Numerical Example 19 3.1.3 IOM’s Three Smoothing Specifications 22 3.1.4 Wage Measurement Recommendations 23

3.2 Recommended Changes to Measurement of Physician Wages 24 3.3 Recommended Changes to Data Sources Used to Compute Office Rents 26 3.4 Endorsement of Current Purchased Services Index Methodology 27 3.5 Endorsement of Current GPCI Cost Share Weights 27

4 Evaluation of GPCI Employee Wage Recommendations 29 4.1 Characterization of IOM’s Recommended Payment Areas for Labor Markets 29

4.1.1 Simple Depiction of Labor Markets for Physician Offices 30 4.1.2 Calculation of IOM Employee Wage Index in this Example 33 4.1.3 Issues with IOM’s Commuting Shares 34 4.1.4 Illustrations of IOM’s Out-Commuting Adjustment 34

4.2 Empirical Impacts of IOM’s Commuting-Based Smoothing Approach 39 4.2.1 Out-Commuting Adjustment’s Effect on the Presence of GAF Cliffs 40 4.2.2 Data Sources for Implementing IOM Commuting Adjustments 41

4.3 Measuring Wages of Workers in the Healthcare Industry 42 4.3.1 Advantages and Disadvantages of Using Industry-Specific Wage Data 42 4.3.2 Advantages and Disadvantages of Using Confidential BLS OES Data 45

5 Evaluation of Physician Work GPCI Recommendations 47 5.1 Discussion of Regression Approach for Predicting Physician Wages 47 5.2 Methods and Data Sources for Measuring Physician Wages 50

5.2.1 IOM Proposed Adjustments of Physician Earnings 50 5.2.2 Candidate Data Sources for Predicting Physician Wages 52

5.3 Empirical Findings Using IOM Regression Approach 56 5.3.1 Regression Specifications 57 5.3.2 Regression Results .... 60

5.4 Estimates Using Alternative Proxy Occupations 66 6 Evaluation of GPCI Office Rent Recommendations 69 6.1 Creating a New Data Source for Commercial Rents 69 6.2 Existing Commercial Rent Data Sources 70

xii Table of Contents Acumen, LLC


............................................................................................................

....................................................................................................................

...................................................................................................................

................................................

...........................................................................

.....................................

........................................................................................

..................................................................................

............................................................................................

...................................................................................

................................................................

.............................................................................

..........................

..............................................................

.......................

..........................

................................

...........................

...........

.........................................................................................................

.............................................

.......................................................

.....................................................

.....................................

....................................................................................................................................

..................................................

........................

..............

...............................

...............

...................................................................

..........................................................

...............................................

...........................................................................................

....................................

....................................

..................................................................

....................................

6.2.1 CoStar Group 71

6.2.2 LoopNet 72 6.2.3 Reis, Inc. 72 6.2.4 Medical Group Management Association (MGMA) 72 6.2.5 Federal Agencies: USPS and GSA 73 6.2.6 Overview: Comparison of Commercial Rent Data Sources 73

6.3 Residential Rent Data Sources 75 6.3.1 American Community Survey 75 6.3.2 HUD Fair Market Rents 76 6.3.3 Basic Allowance for Housing 77

7 Empirical Impacts of IOM Recommendations 78 7.1 Effects of Redefining Payment Areas 78

7.1.1 Implementing the Out-Commuting Based Smoothing Adjustment 79 7.1.2 Payment Area Definitions Based on MSAs 81 7.1.3 Implementing IOM Three-Tiered Payment Area Recommendations 85

7.2 Effects of Regression-Based Methodology for Calculating PW GPCI 86 7.3 Effects Using an Alternative Set of PW GPCI Proxy Occupations 88

7.3.1 Impacts of Alternative Occupations Using Current Methodology 89 7.3.2 Impacts of Alternative Occupations Using Regression-Based Methodology 90

8 Summary of Findings 93 8.1 Evaluation of IOM’s Employee Wage Recommendations 93 8.2 Evaluation of IOM’s PW GPCI Recommendations 94 8.3 Evaluation of IOM’s Office Rent Recommendations 96 8.4 Empirical Impacts of IOM Recommendations on GAF Values 96

References 98 Appendix A : Current Employee Wage Index Calculation 101

A.1 Selecting the occupations for inclusion in the wage index calculation 101 A.2 Calculating an RVU-weighted national average hourly wage by occupation 101 A.3 Indexing the occupation wage in each MSA to the national wage 102 A.4 Calculating occupations’ share of the national employee wage expenditure 102 A.5 Calculating MSA-level hourly wage index 102 A.6 Calculating locality-level employee wage index 103

Appendix B : Impact of IOM’s MSA-Based Payment Areas 104 B.1 Payment Areas Definitions 104 B.2 Measuring Variability Across Four Candidate Payment Areas 105 B.3 Measuring Variability Across Four Candidate Payment Areas 106

Appendix C : Unweighted PW GPCI Regression 110 Appendix D : Alternative Specification for the Proxy Occupations 112

xiii

..........................................................

..................................

..................

......................................

..........................................................

.....................................................

.......................................................

................................................................

.............................................................

.....................................................................................

...............

.............................

.............................

......

........................................................

..............................

..............................................

..................................................

...................................

...................................

.......

......................

.........................

.............................................................

.......

...................................................

...........................................................

.....................

.........................

....

..........................................

.............................

..............................................................

...............................................

..........................

................................................

........................................................

.............................

.........................

..............

.....................................................

...........

..........................................

..................................................

LIST OF TABLES AND FIGURES

Table 1: Breakdown of GPCIs into Six Component Indices ii Table 2: IOM Geographic Practice Cost Index (GPCI) Recommendations iii Table 3: IOM’s Suggested Three-Tiered System for Defining GPCI Payment Areas iv Table 4: Example Application of the IOM Out-Commuting Adjustment v Table 5: Summary of Changes to PW GPCI Components viii Table 6: Distribution of Changes in GAF for Impact Analyses x Table 2.1: Breakdown of GPCIs into Six Component Indices 6 Table 2.2 Wage Bill Shares for Fifth and Sixth Updates 8 Table 2.3: Data Sources Used for Recent GPCI Updates 12 Table 3.1: IOM GPCI Recommendations 14 Table 3.2 IOM’s Suggested Three-Tiered System for Defining GPCI Payment Areas 15 Table 3.3: Illustrating Step 1 of the IOM Employee Wage Index Calculation 20 Table 3.4: Illustrating Step 2 of the IOM Employee Wage Index Calculation 21 Table 3.5: Application of Smoothing Adjustments under Three IOM Outmigration Models 22 Table 3.6: Summary of Changes to PW GPCI Components 24 Figure 4.1: Illustration of Local Labor Markets for Three Physician Offices 31 Table 4.1: Commuting Shares and Wages, Urban-Rural Example 35 Table 4.2: Calculation of the IOM EWI, Urban-Rural Example 36 Figure 4.2: Illustration of Local Labor Markets with Commuting Barrier 38 Table 4.3: Commuting Shares and Wages, Commuting-Barrier Example 38 Table 4.4: Counterintuitive Implication of IOM Smoothing, Commuting-Barrier Example 38 Figure 4.3: Difference in County GAF Values with Out-Commuting Adjustment 41 Table 4.5: Example of Cross-Industry Wage Variability for Registered Nurses 44 Table 4.6: Nursing Wages by Industry (BLS OES 2010) 44 Table 4.7: Concentration of Physicians’ Workers in Healthcare Industry (BLS OES 2010) 45 Table 5.1: Proposed Adjustments for Physician Earnings Data 51 Table 5.2: Data Available to Measure Physician Earnings 53 Table 5.3: Method of Physician Compensation by Specialty (2011 MGMA Data) 54 Table 5.4: Regression Results for PW GPCI Using Current Proxy Occupations 61 Table 5.5: Family and General Practitioner Wages by Rural-Urban Status (BLS OES 2008) 63 Table 5.6: Regression Results for PW GPCI Using 2005-2009 ACS 65 Table 5.7: Regression Results for PW GPCI Using Alternative Occupations 68 Table 5.8: Comparison of PW GPCI Regressions Using Original and Alternative Occupations 68 Table 6.1: Comparison of Data Sources for Office Rent 74 Table 7.1: Summary of IOM’s Payment Area Recommendations 79 Table 7.2: Difference in Employee Wage Index With and Without Smoothing 80 Table 7.3: Difference in PE GPCI With and Without Smoothing 81 Table 7.4: Difference in GAF With vs. Without Smoothing 81

Table 7.5: Difference in GAF when Switching to MSAs (Locality Baseline) 83Table 7.6: Difference in GAF when Switching to Counties (Locality Baseline) 83 Table 7.7: Difference in GAF when Switching to Statewide Tiers (Locality Baseline) 83 Table 7.8: Change in GAF by Urban-Rural Continuum Code 85 Table 7.9: Difference in GAF: IOM Three-Tiered Payment Area vs. Medicare Locality 86 Table 7.10: Impact Analysis: Specialty-Mix Regression (PW GPCI) 87 Table 7.11: Impact Analysis: Specialty-Mix Regression (GAF) 87

xiv Table of Contents Acumen, LLC


.......................................

...............................................

...................................

...........................................

...

...........

......................

..............................

...........................................

..................

..............

..............

............

..........................

...........................

Table 7.12: Impact Analysis: Family Practice Regression (PW GPCI) 88

Table 7.13: Impact Analysis: Family Practice Regression (GAF) 88 Table 7.14: Alternative Proxy Occupations Impact Analysis (PW GPCI) 89 Table 7.15: Alternative Proxy Occupations Impact Analysis (GAF) 90 Table 7.16: Alternative Occupation Impact Analysis: Specialty-Mix Regression (PW GPCI) 91 Table 7.17: Alternative Occupation Impact Analysis: Specialty-Mix Regression (GAF) 91 Table 7.18: Alternative Occupation Impact Analysis: FP Regression (PW GPCI) 91 Table 7.19: Alternative Occupation Impact Analysis: FP Regression (GAF) 92 Table B.1: Summary of GAF Values by Alternate Payment Areas 106 Figure B.1: Difference in County GAF Values in Different Localities by Distance 108 Figure B.2: Share of Counties with GAF Differential Greater Than 0.05 by Distance 109 Table C.1: Weighted vs. Unweighted PW GPCI Regressions (Original Occupations) 110 Table C.2: Weighted vs. Unweighted PW GPCI Regressions (Alternate Occupations) 111 Table D.1: Summary Statistics for Alternative PW GPCI Proxy Occupations 112 Table D.2: Regression Results for PW GPCI Using Alternative Occupations 114

xv

LIST OF ABBREVIATIONS

ACA: Affordable Care Act

ACO: Accountable Care Organization

ACS: American Community Survey

ACS PUMS: American Community Survey Public Use Micro Sample

AMA: American Medical Association

AMA PPIS: American Medical Association Physician Practices Information Survey

BLS: Bureau of Labor Statistics

BOMA: Building Owners and Managers Association

CF: Conversion Factor

CMS: Centers for Medicare and Medicaid Services

CPI: Consumer Price Index

CTPP: Census Transportation Planning Package

CTS: Community Tracking Survey

CWA: Census Work Area

CY: Calendar Year

DOD: United States Department of Defense

DOD BAH: U.S. Department of Defense Basic Allowance for Housing

ECI: Employment Cost Index

EWI: Employee Wage Index

FMR: US Department of Housing and Urban Development’s Fair Market Rent data

FDIC: Federal Deposit Insurance Corporation

FS: Full Service

FY: Fiscal Year

GAF: Geographic Adjustment Factor

GAO: Government Accountability Office

GPCI: Geographic Practice Cost Index

GSA: General Services Administration

HCPCS: Healthcare Common Procedure Coding System

HSPA: Health Professional Shortage Areas

HUD: United States Department of Housing and Urban Development

HWI: Hospital Wage Index

IOM: Institute of Medicine

IPPS: Inpatient Prospective Payment System

LPN: Licensed Practical Nurse

MEI: Medicare Economic Index

MG: Modified Gross

MGMA: Medical Group Management Association

MHA: Military Housing Area

MHW: Mean/Median Hourly Wage

MP: Malpractice

MSA: Metropolitan Statistical Area

OACT: Office of the Actuary

OB/GYN: Obstetrician/Gynecologist

OES: Occupational Employment Statistics

PE: Practice Expense

PFS: Physician Fee Schedule

PW: Physician Work

RBRVS: Resource-Based Relative Value Scale

RN: Registered Nurse

RVU: Relative Value Unit

SGR: Sustainable Growth Rate

USDA: United States Department of Agriculture

USPS: United States Postal Service

xvi Table of Contents Acumen, LLC

1 INTRODUCTION

Medicare pays physicians for their services according to the Physician Fee Schedule

(PFS), which specifies a set of allowable procedures and payments for each service. Each

procedure is interpreted as being produced by a combination of three categories of inputs:

physician work (PW), practice expense (PE), and malpractice insurance (MP). The particular

blend of PW, PE, and MP inputs assessed to produce a service specifies its composition of

relative value units (RVUs). A payment for a procedure depends on its assigned RVUs and the

input prices assessed for each RVU component. Under mandates in Section 1848(e) of the

Social Security Act, the Centers for Medicare and Medicaid Services (CMS) must apply

geographic cost indices in the calculation of component RVU input prices. Starting in 1992,

CMS introduced Geographic Practice Cost Indices (GPCIs) to comply with this mandate; CMS

updates GPCIs at least every three years.

Concerns have been expressed regarding the accuracy of GPCIs in measuring physicians’

regional cost differences. In a 2005 report, the Government Accountability Office (GAO) stated

that the "geographic adjustment indices are valid in design," but questioned the applicability of

the wage and rental data used to calculate PE GPCIs.1

1 U.S. GAO March 2005.

GAO recommended augmenting wage

data to cover a wider array of occupations and basing rents on commercial office rents instead of

residential rents which GPCIs currently rely upon. GAO also advised CMS to refine malpractice

GPCIs by standardizing input data collection and making them more complete and

representative. In addition to changes in the wage, rent, and malpractice premium data CMS

uses, GAO further raised issues about how to measure physician wages when some physicians

are self-employed and other are salaried, as well as what area is applicable for defining physician

wage indices. GAO, along with other critics, have questioned the appropriate constructions of

localities for calculating all forms of GPCIs; GAO found that substantial variation in practice

costs existed within each payment area under the current locality-based system.

In its latest efforts to improve the methodology and data sources used to compute GPCIs

and other geographic input cost adjustments, CMS sponsored the Institute of Medicine (IOM) to

produce a series of reports examining how CMS measures geographic variation in input prices

faced by physicians.2

2 In addition to GPCIs, IOM was also asked to evaluate the Hospital Wage Index (HWI) methodology used by CMS

to adjust payments to hospitals and other institutional providers.

In its Phase I report published in September 2011, IOM’s "Committee on

Geographic Adjustment Factors in Medicare Payment" evaluates the accuracy of the current

geographic adjustment factors, the methodology used to make adjustments, and the extent to

which alternative sources of data are representative of relevant circumstances for healthcare

providers. The IOM report offers a range of recommended modifications to the methodology

Acumen, LLC Geographic Adjustment of Medicare Payments to Physicians 1

and data used to compute the hospital wage index (HWI) and GPCIs.3

3 IOM 2011.

Regarding GPCIs, some

of IOM’s recommendations support CMS’s current practices (e.g., continued use of the MEI cost

share weights) and others that CMS has already adopted for calendar year (CY) 2012 (e.g.,

creation of the purchased service index).

The new changes to the GPCI calculations recommended by IOM fall principally into

three categories of modifications in methodologies and data:

(1) Compute the employee wage components of the PE GPCI using counties as payment

areas with wages adjusted for commuting patterns and using data on healthcare

workers;

(2) Use a regression-based approach to measure regional variation in physician wages in

the PW GPCI; and

(3) Identify a source of commercial office rent data to measure regional variation in

physicians’ cost to rent office space as part of the PE GPCI.

IOM recommendation (1) argues for redefining payment areas for employee wage indices as the

county in which a physician office is located with wages measured to account for workers’

commuting patterns across metropolitan statistical areas (MSAs) and with wage data on workers

from firms in the healthcare industry (rather than from all industries) recognizing occupational

mixes consistent with workforces in physician offices. This revision of the PE GPCI wage

component would align it with IOM’s recommendations regarding calculation of wage indices

for hospitals and other institutional providers. IOM recommendation (2) would replace CMS’s

current PW GPCI values, which are equal to a weighted average of proxy-occupation wage index

values, with a regression framework to compute regional differentials in physician wages.

Finally, IOM recommendation (3) suggest replacing the residential rent data currently used to

measure regional variation in office rents with a new source of office rent data.

The discussion in the sections below evaluates the potential impacts of implementing the

IOM recommendations from both conceptual and empirical perspectives. The conceptual

analysis weighs the advantages and disadvantages of each of the three IOM recommendations

categories listed above. Additional work evaluates alternative methods for formulating payment

areas and labor markets across multiple GPCI component indices. The empirical analysis

investigates whether the identified conceptual challenges become problematic in practice, and it

further explores the impacts of the IOM recommendations on the values of GPCI indices.

The remainder of this report consists of seven sections. Section 2 provides an overview

of the Resource-Based Relative Value Scale (RBRVS) system and describes how CMS currently

uses GPCIs to adjust physician payments. Section 3 explains IOM’s recommended changes to

2 Introduction Acumen, LLC

the GPCI methodology. Sections 4, 5, and 6 evaluate each of the three IOM recommendation

categories described above in detail. Specifically, Section 4 examines issues related to

measuring regional variation in employee wages, Section 5 evaluates IOM’s proposals to

redefine the methodology used to measure regional variation in physician wages, and Section 6

assess potential sources of office rent data that CMS could use to calculate the office rent index.

Section 7 presents an empirical analysis showing the prospective impacts of adopting IOM

recommendation on the values of GPCIs. Finally, Section 8 concludes with a summary of

findings.


2 GEOGRAPHIC ADJUSTMENTS OF PHYSICIAN FEE SCHEDULE UNDER CURRENT POLICY

Where physicians locate their practices affects their cost of providing each service. For

instance, the cost of living for physicians is higher in Manhattan than in Montana; the cost of

operating a physician practice is higher in San Francisco than in Sandusky, Ohio; and purchasing

malpractice insurance is more expensive for a physician in Miami than for one in Minneapolis.

To account for these geographic differences in input costs, CMS modifies the payments it makes

to physicians using GPCIs. GPCIs adjust physician payments based on geographic differences in

physician wages, practice expenses, and the price of malpractice insurance. In fact, CMS creates

three GPCIs—PW, PE, and MP—which correspond to the three broad classes of inputs

physician practices use.

The remainder of this section provides additional background information regarding how

CMS uses GPCIs within the Medicare PFS. Specifically, this section answers three questions:

How do GPCIs affect Medicare payments to physicians?

What are the six component indices that make up GPCIs?

What methodology does CMS currently use to calculate GPCIs?

The following three sections answer each of these questions in the order they appear above.

2.1 How GPCIs Affect Physician Payments

Under the PFS, Medicare pays for physician services based on a list of services and their

payment rates. Under the PFS, every physician service corresponds to a specific procedure code

within the Healthcare Common Procedure Coding System (HCPCS). Since 1992, CMS has

relied on the RBRVS system to determine the fee for each procedure. In the RBRVS system,

payments for each service depend on the relative amounts of inputs required to perform the

procedure. These inputs include the amount of physician work needed to provide a medical

service, expenses related to maintaining a practice, and malpractice insurance costs. CMS

estimates the quantity of inputs required to provide these services using PW, PE, and MP RVUs,

respectively.

The three GPCIs adjust their corresponding RVUs for regional variation in the price of

each of the three input categories. GPCIs increase the RVU values for high-cost areas and

reduce the RVU values for low-cost areas. GPCIs do not affect aggregate payment levels;

instead, they reallocate payment rates by locality to reflect regional variation in relative input

prices. For instance, a PE GPCI of 1.2 indicates that practices expenses in that area are 20

percent above the national average, whereas a PE GPCI of 0.8 indicates that practices expenses

in that area are 20 percent below the national average.

4 Geographic Adjustments of Physician Fee Schedule Under Current Policy Acumen, LLC

CMS calculates the three GPCIs for 89 payment areas known as Medicare localities.

Each physician payment locality is assigned an index value, which equals input cost estimates

within each payment area over the average input cost at the national level. Localities are defined

alternatively by state boundaries (e.g., Wisconsin), MSAs (e.g., Metropolitan St. Louis, MO),

portions of an MSA (e.g., Manhattan), or rest-of-state area which exclude metropolitan areas

(e.g., Rest of Missouri).4

4 An MSA is comprised of one or more counties and includes the counties that contain a core urban area with a

population of 50,000 or more, as well as surrounding counties that exhibit a high degree of social and economic

integration. For more information, see the U.S. Census Bureau website: http://www.census.gov/population/metro/.

As a result, some localities are large metropolitan areas, such as San

Francisco and Boston, whereas many are statewide payment areas that include both metropolitan

and nonmetropolitan areas, such as Minnesota, Ohio, and Virginia.5

5 For a brief history of the changes to GPCI payment areas from their inception in 1966 to the current regulation,

see: U.S. GAO June 2007 and CMS 1993.

Using the RVUs, GPCIs, and a conversion factor (CF), one can calculate the physician

payment for any service in any locality. The CF translates the sum of the GPCI-adjusted RVUs

into a payment amount. Equation (2.1) below demonstrates how the PW, PE, and MP GPCIs

combine with the three RVUs and the CF to establish a Medicare physician payment for any

service H in locality L:6

6 The Medicare physician payment calculated using equation (2.1) may also be adjusted upwards or downwards

through payment modifiers. For example, physicians use a modifier to bill for a service when they assist in a

surgery; payment for an assistant surgeon is only a percentage of the fee schedule amount for the primary surgeon.

(2.1)

Although GPCIs affect payments for each procedure depending on the relative amounts

of PW, PE, and MP RVUs, one can summarize the combined impact of the three GPCI

components on a locality’s physician reimbursement levels using the Geographic Adjustment

Factor (GAF). The GAF is a weighted average of the three GPCIs for each locality, where the

weights are determined by the Medicare Economic Index (MEI) base year weights. Using the

2006 MEI base weights, one can calculate the GAF as follows:

(2.2)

2.2 GPCIs’ Six Component Indices

CMS uses six component indices to calculate the three GPCIs. Table 2.1 maps the

corresponding component index to its relevant GPCI. Whereas the PW and MP GPCIs are

comprised of a single index, the PE GPCI is comprised of four component indices (i.e., the

employee wage; purchased services; office rent; and equipment, supplies and other indices). The

first component of the PE GPCI, the employee wage index, measures regional variation in the

cost of hiring skilled and unskilled labor directly employed by the practice. Practice expenses


http://www.census.gov/population/metro/

Table 2.1: Breakdown of GPCIs into Six Component Indices

for employee wages account for the largest share of the PE GPCI. Although the employee wage

index adjusts for regional variation in the cost of labor employed directly by physician practices,

the employee wage index does not account for geographic variation of practices’ costs for

services that have been outsourced to other firms. Such cases occur when practices purchase

services from law firms, accounting firms, information technology consultants, building service

managers, or any other third-party vendor. The second component, the purchased services index,

measures regional variation in the cost of these contracted services that physicians typically buy.

The third component, the office rent index, measures regional variation in the cost of typical

physician office rents. For example, renting an office in San Francisco is more expensive than

renting an office in Wyoming; the office rent index produces an estimate of this regional

variation in the price of office space. Finally, the "equipment, supplies and other" category

measures practice expenses associated with a wide range of costs from chemicals and rubber, to

telephone and postage. CMS assumes that these capital goods are purchased in a national market

and does not adjust for regional variation in practice costs within the "equipment, supplies and

other" category; thus, each locality receives a value of one for the "equipment, supplies and

other" index.

GPCI Component Index Measures Geographic Differences in:

Physician

Work Single Component Physician wages

Practice

Expense

Employee Wage Wages of clinical and administrative office staff

Purchased Services Cost of contracted services (e.g., accounting, legal,

advertising, consulting, landscaping)

Office Rent Physician cost to rent office space

Equipment, Supplies, and Other Practice expenses for inputs such as chemicals and

rubber, telephone use and postage

Malpractice Single Component Cost of professional liability insurance

2.3 Current Policy for Calculating GPCIs

Calculating GPCI values requires measuring the price of each input relative to its national

average price. Although the general approach is similar across all geographically-adjusted

component indices, the specific methodology and data used to calculate each index value vary.

For instance, whereas the employee wage index measures worker wages directly, the PW GPCI

measures regional variation in physician wages using proxy occupations; whereas labor-related

indices rely on wage data from the Bureau of Labor Statistics (BLS) Occupational Employment


Statistics (OES); the office rent index uses the American Community Survey (ACS) to measure

regional variation in office rents.

The remainder of this section describes the methodology for calculating the six GPCI

component indices. Sections 2.3.1, 2.3.2 and 2.3.3 contain an overview of the methodology for

calculating the component indices within the PW GPCI, PE GPCI, and MP GPCI, respectively.

Section 2.3.4 describes the data CMS currently uses to calculate each GPCI component. Section

2.3.5 presents some of the legislative adjustments that affect GPCI values but which are not

discussed in the general GPCI methodology. A more detailed description of the methodology

used to calculate the GPCI component indices can be found in previous reports describing the

Sixth Update7

7 O’Brien-Strain, et al. November 2010.

and Revisions to the Sixth Update.8

8 MaCurdy, et al. October 2011.

2.3.1 Physician Work GPCI Methodology

In the current methodology, CMS defines PW GPCI values based on regional variation in

wages across a set of proxy occupations. Although direct measures of physician wages are

available in nationally representative data sources (e.g., BLS OES, ACS), CMS elects not to use

this information in its PW GPCI calculation. According to a 2005 GAO report, computing the

PW GPCI using direct measures of physician wages would produce a circular measure where the

work adjustment would depend on past payments to physicians by Medicare; to attenuate this

problem, CMS uses proxy occupation wages in its calculation of PW GPCI values. Specifically,

CMS uses the following four steps to calculate the PW GPCI:

(1) Select proxy occupations and calculate an occupation-specific index for each proxy;

(2) Assign weights to each proxy-occupation index to create an aggregate proxy-

occupation index at the locality level;

(3) Adjust the aggregate proxy-occupation index by a physician inclusion factor; and

(4) Re-scale the PW GPCI to ensure budget neutrality.

The proxy occupations Medicare currently selects in the first step represent highly

educated, professional occupation categories, whose wages would be expected to reflect the

overall geographic differences in living costs and amenities for other professional workers. To

develop a labor cost index for the physician’s own work, the current PW GPCI draws on the

regional variation in the earnings of the following professionals:

Architecture and Engineering,

Computer, Mathematical, Life and Physical Science,

Social Science, Community and Social Service, and Legal,

Education, Training, and Library,


Registered Nurses,

Pharmacists, and

Art, Design, Entertainment, Sports, and Media.

Using BLS OES data, CMS calculates an occupation-specific index for each of the proxy groups.

The occupation-specific index in a given county is the median hourly earnings for that

occupation relative to RVU-weighted national average median hourly earnings. As BLS OES

wage data are reported by MSA, all counties in the same MSA receive the same proxy

occupation index value.

To create an aggregate proxy-occupation index, the second step weights these

occupation-specific indices by each occupational group’s share of the national wage bill. An

occupation’s share of the national wage bill equals the national hourly wage for that occupation

multiplied by the number of non-zero wage earners in that occupation nationally and then

divided by the wage bill summed across all proxy occupations. Table 2.2 lists the wage bill

shares utilized in the Fifth and Sixth Updates for the seven occupation groups.

Table 2.2 Wage Bill Shares for Fifth and Sixth Updates

Occupation Group Fifth Update Sixth Update

Architecture and Engineering 13.9% 8.5%

Computer, mathematical, life and

physical science 19.1% 16.0%

Social science, community & social service, and legal

15.5% 8.5%

Education, training, and library 30.6% 40.2%

Registered nurses 11.1% 16.6%

Pharmacists 1.6% 2.8%

Art, design, entertainment, sports,

and media. 8.2% 7.4%

Total 100% 100%

Using the wage bill share, one can calculate the county-specific hourly index as the sum

of the product of the county indices for each occupation times the wage bill share for each

occupation. The preliminary county-level physician wage index is then aggregated to the

locality level by weighting the county indices described above by the number of PW RVUs in

each county. Then, one can translate the county-level PW GPCI index to a locality-level index

using the following formula:

(2.3) 𝑋𝐿 = 𝑅𝑉𝑈𝑃𝐸 ,𝑘 × 𝑋𝑘𝑘∈{𝑘𝐿}

𝑅𝑉𝑈𝑃𝐸 ,𝑘𝑘∈{𝑘𝐿}

where XL is the locality-level index composite index, Xk is the county-level index, and RVUPE,k is

the number or PE RVUs that were billed in each county. The expression 𝑘 ∈ {𝑘𝐿} indicates the

summation over all counties that are located in locality L.


The third step implements the Congressionally-mandated PW GPCI inclusion factor. The

inclusion factor reduces the magnitude of the variability in the PW GPCI. After applying the

physician inclusion factor, the adjusted PW GPCI can be calculated as:

(2.4) 𝐺𝑃𝐶𝐼𝑃𝑊,𝐿 = 1 + 𝐼𝑛𝑐𝑙𝑢𝑠𝑖𝑜𝑛 𝐹𝑎𝑐𝑡𝑜𝑟 × 𝑋𝐿 − 1

where the left hand side variable is the PW GPCI for locality L, and XL is the locality proxy

estimated in the second step above. An inclusion factor of one (i.e., 100 percent) would account

for all observable variation in physician wages, and the PW GPCI would equal the locality proxy

XL; an inclusion factor of zero (i.e., 0 percent) would remove geographic adjustments and would

set the PW GPCI to one in all areas. As mandated by section 1848(e)(1)(A)(iii) of the Social

Security Act, the current inclusion factor is 25 percent. If the locality proxy was 1.4, for

example, after applying the 25 percent inclusion factor the PW GPCI would equal 1.1. Reducing

the inclusion factor aims to equalize physician compensation across areas.9

9 Zuckerman et al. September 2004.

The fourth and final step rescales the PW GPCI to ensure budget neutrality. Budget

neutrality adjustments are applied in the final step of calculating each GPCI to ensure that the

total payments distributed remain the same under the updated PW GPCIs as they were under the

previous PW GPCIs.

2.3.2 Practice Expense GPCI Methodology

Although the approach for calculating each of the four PE GPCI component indices

differs, all geographically-adjusted indices broadly follow the same three steps. To present the

general framework for calculating the PE GPCI indices, this section begins by describing the

approach for the office rent index, which uses the following steps:

(1) Calculate an RVU-weighted national average rent value using county rent data;

(2) Create a county-specific index; and

(3) Calculate a Medicare locality-level index.

The office rent index currently measures regional variation in the price of office rents

using residential rent data from the ACS on median gross rents for two-bedroom apartments. In

step 1, one calculates national average rents as follows:

(2.5) 𝑅𝑁 = 𝑅𝑉𝑈𝑃𝐸 ,𝑘 × 𝑅𝑘𝑘

𝑅𝑉𝑈𝑃𝐸 ,𝑘𝑘

where RN is the RVU-weighted national average, RVUPE,k is the number of PE RVUs in county k,

and Rk is the median gross rent in county k. Using the national rent estimate, one can create a

county-specific rent index in step 2 as the ratio of the county gross rents and the national average

rents as follows:


(2.6) 𝑋𝑘 =𝑅𝑘𝑅𝑁

In this case, Xk is the office rent index for county k. In step 3, one aggregates the county-level

office rent index to locality-level office rent index as shown in equation (2.3).

Although the employee wage index relies on a similar approach, CMS relies on wage

data across multiple occupations to create a composite index describing regional variation in the

wages of workers typically employed by physician practices. To compute a composite index for

any county, one follows the same steps used to compute the PW GPCI with the exception that no

inclusion factor is applied (or, equivalently, the inclusion factor is 100%). When translating this

approach to the employee wage index case, step 1 creates a county-level index for each

occupation employed in the offices of physician industry, where the county-level occupation

specific index equals the occupation’s median wage in the county divided by the RVU-weighted

national average wage for that occupation. Unlike the PW GPCI, the employee wage index

directly measures the wages of workers employed by physicians and does not use proxy

occupations. Step 2 calculates a composite wage index for each county as a weighted average of

these occupation-specific indices. The weights in this weighted average equal each occupation’s

share of the national wage bill within the offices of physicians industry. Once CMS calculates

the composite wage for each county, one aggregates the county-level index to the locality level

as described in equation (2.3).

The methodology for computing the purchased services index follows the same broad

approach with three modifications. First, rather than including occupations that are employed in

physician offices, the purchased services index includes occupations employed in industries from

which physicians are likely to purchase services. Second, the weight each occupation receives in

the composite index differs between the employee wage index and purchased services index.

Whereas the employee wage index weights each occupation based on each share of the national

wage bill in the offices of physician industry, the purchased services index weights occupations

based on their national wage share within the industries from which physicians purchase

services. Third, unlike the employee wage index, only a portion of the purchased services index

is geographically adjusted. Because capital expenses make up approximately 38 percent of

purchased services inputs, only 62 percent of the index is adjusted for regional variation in labor

costs. 10

10 The exact proportion of the occupation-specific index that is regionally adjusted depends on the labor-related

share of expenses in the industries in which that occupation is most frequently employed.

The only PE GPCI component that does not follow the general methodology presented

above is the "equipment, supplies and other" index. This index is not geographically adjusted.

Thus, all localities receive an equipment and supplies component index value of 1.0.


2.3.3 Malpractice GPCI Methodology

MP GPCI largely follows the general PE GPCI methodology but has three unique

features. First, like the employee wage index, the MP GPCI is a composite index; whereas the

employee wage index is a composite of median wages for specific occupations, however, the

malpractice GPCI is a composite index that combines measures of regional variation in

malpractice premiums across physician specialties. To create the specialty-mix adjusted

composite index, one calculates a county-specific index based on the premium levels for each

specialty, and then one calculates the composite county-index as a weighted average of these

specialty-specific malpractice indices. Second, whereas all PE GPCI component indices use

national weights when creating a composite index, the malpractice GPCI relies on state-specific

specialty weights. This specification reflects the fact that state malpractice premiums by

specialty in part reflect the norms of care in each state. Third, whereas most other component

indices use ACS or BLS data to create their index values, CMS principally uses malpractice

premium state rate filing data.11

11 For a detailed description of the malpractice premium data used for the MP GPCI, see O’Brien Strain et al.

November 2010.

2.3.4 Data Sources Used to Calculate GPCIs

CMS relies on a number of data sources to calculate the GPCI components. Table 2.3

compares the data sources used under the 2012 Sixth Update and the Revisions to the Sixth

Update implemented in CY 2012. Of particular importance are the BLS OES establishment data

and the ACS household data. CMS uses the former to measure regional variation in the cost of

labor-related inputs and the latter to measure regional variation in rents.

The BLS OES survey is a semi-annual mail survey of all salaried non-farm workers,

excluding self-employed individuals, administered by the BLS. OES data from any year are

aggregated using six semi-annual panels collected over three years.12

12 The BLS OES uses data over time to increase the sample size of the survey, thereby increasing reliability and

reducing sampling error. But labor costs change over time, as evidenced by the Employment Cost Index (ECI) time

series data. To make the data from all survey respondents comparable, the OES program uses the ECI to translate

the occupation-level wages from previous years into a wage number for the most recent year. For additional details,

see the BLS OES Technical Notes: http://www.bls.gov/oes/current/oes_tec.htm.

The 2008 OES wage

estimates, for example, contain employer survey responses from May 2008, November 2007,

May 2007, November 2006, May 2006, and November 2005. The establishments surveyed are

selected from lists maintained by State Workforce Agencies for unemployment insurance

purposes. To create a sample for the OES data, BLS selects establishments from every

metropolitan area and state, across all surveyed industries, and from establishments of varying

sizes. The OES program produces employment and wage estimates for over 800 occupations

across 23 major occupational groups, including "healthcare practitioners" and "healthcare


http://www.bls.gov/oes/current/oes_tec.htm

support occupations." Using this sample of establishments, the BLS collects detailed wage data

by industry, occupation, and region. For instance, the BLS OES data contain industry wage

information for the healthcare sector and the offices of physicians industry.

Table 2.3: Data Sources Used for Recent GPCI Updates

Component

Sixth Update

2012

Revisions to the Sixth Update

2012 (Current Regulation)

Physician Work GPCI 2006-2008 BLS Occupational

Employment Statistics

2006-2008 BLS Occupational


Practice Expense

GPCI

Employee Wage 2006-2008 BLS Occupational




Office Rent FY2010 HUD 50th

Percentile Rents

2006-2008 American Community

Survey

Purchased Services

(Labor Cost) N/A



Purchased Services

(Labor Related Shares) N/A CMS Labor-Related Classification

Equipment, Supplies, Other 1.000 for all counties 1.000 for all counties

Malpractice GPCI 2006-2007

Malpractice Premiums

2006-2007

Malpractice Premiums

Cost Share Weights 2000 MEI weights 2006 MEI weights

County RVU Weights 2008 RVUs 2009 RVUs

To estimate prevailing rental costs, CMS uses 2-bedroom rental data from the 2006-2008

American Community Survey. The ACS is an annual household survey conducted by the U.S.

Census Bureau. The ACS samples nearly 3 million addresses each year, resulting in nearly 2

million final interviews, and replaces the decennial census long form.13

13 U.S. Census Bureau November 2008.

To calculate the office

rent index, CMS relies on a customized extract of the ACS data to measure average gross rents

for each county.14

14 Utilities cannot be analyzed separately since some individuals’ monthly rent covers the cost of utilities. Thus the

2006-2008 ACS data can only accurately measure gross rents (i.e., including utilities) rather than net rents.

For counties with fewer than 20,000 individuals, however, ACS does not

publicly release rental rate data.

2.3.5 Legislative Adjustments

CMS implements a number of required adjustments after completing the core GPCI

calculations. Section 1848(e)(1)(E) of the Act provides for a 1.0 floor for the PW GPCI, which

was set to expire at the end of 2011, until it was extended through the end of CY 2012 by the

Temporary Payroll Tax Cut Continuation Act of 2011 and the Middle Class Tax Relief and Job

Creation Act of 2012. In addition, Section 1848(e)(1)(G) of the Social Security Act sets a


permanent 1.5 PW GPCI floor for services furnished in Alaska beginning January 1, 2009.

Further, section 1848(e)(1)(I) establishes a 1.0 PE GPCI floor for physicians' services furnished

in frontier States effective January 1, 2011. The following states are considered to be "Frontier

States" for CY 2013: Montana, North Dakota, Nevada, South Dakota, and Wyoming. The

empirical analyses in this report, however, detail only the calculations of GPCIs before final

adjustments.


3 DESCRIPTION OF IOM’S GPCI RECOMMENDATIONS

IOM recommended alterations of GPCIs fall into five broad categories. Table 3.1 maps

each of IOM’s recommendations to the associated category and provides a brief description of

each recommendation. The first category includes IOM proposals related to calculation of the

employee wage components of the PE GPCI, which suggest using counties as payment areas

with wages adjusted for commuting patterns and using data on healthcare workers. The second

category involves replacing CMS’s current use of a weighted average of proxy-occupation wages

by a regression framework to compute regional differentials in the physician wage component of

GPCI. The third category includes recommended improvements in the source of office rent data

that CMS uses to measure regional variation in physicians’ cost to rent office space. The fourth

and fifth categories comprise IOM recommendations that largely mirror modifications already

incorporated in the Revision to the Sixth Update of the GPCI; in particular, the creation of the

purchased service index has been implemented for the FY 2012 GPCIs, and GPCI calculations

continue to use MEI cost share weights which was recently adopted in previous years.

Table 3.1: IOM GPCI Recommendations

Category Number Description

Employee

Wages

2-1 The same labor market definition should be used for both the hospital wage index and the

physician geographic adjustment factor. Metropolitan statistical areas and statewide non-

metropolitan statistical areas should serve as the basis for defining these labor markets.

2-2 The data used to construct the hospital wage index and the physician geographic

adjustment factor should come from all healthcare employers.

4-1

Wage indexes should be adjusted using formulas based on commuting patterns for

healthcare workers who reside in a county located in one labor market but commute to

work in a county located in another labor market.

5-4 The practice expense GPCI should be constructed with the range of occupations

employed in physicians’ offices, each with a fixed national weight based on the hours of

each occupation employed in physicians’ offices nationwide.

5-5 The Centers for Medicare and Medicaid Services and the Bureau of Labor Statistics

should develop a data use agreement allowing the Bu

Geographic Adjustment of Medicare Payments to Physicians ... Adjustment of Medicare Payments to...

Documents

Transcript of Geographic Adjustment of Medicare Payments to Physicians ... Adjustment of Medicare Payments to...